1 March 2019 in Geophysics
Locating microearthquake events below complex heterogeneous overburden requires robust location methodologies that can honor multipathing in the seismic wavefield. We have developed two full-waveform event location methods that form a complementary solution for locating earthquakes and simultaneously deriving focal mechanisms via moment tensor inversion. The methods are based on the application of 3D elastic wavefield modeling, which is used to generate waveforms and extract wavefield attributes, for comparison to the observed field data. Events are located and focal mechanisms are derived via a multiparameter inversion, which minimizes the differences between synthetic and observed data. The results have been applied to the induced seismicity observed within the giant Groningen gas field, onshore Netherlands, where recorded earthquakes are triggered by stress changes, induced in the reservoir through pressure depletion. Locating events below the field is compounded by the presence of strong guided waves, which are trapped in the lower velocity reservoir interval. This complex multivalued wavefield is problematic for traditional event location methods, which assume a single traveltime arrival. We overcome this limitation by using all event arrivals in a wave-based solution to improve the accuracy of locating earthquakes and overcome the ambiguity of solving for location and the focal mechanism simultaneously. The event location methods have been applied to shallow and deep monitoring networks, and 150 events have been located with high accuracy. The interpretation of the earthquake activity indicates that the events studied originate from the movement of larger graben bounding faults, which are oriented in a north-northwest–south-southeast direction.
1 February 2018 in The Leading Edge
Induced seismicity from gas production has been identified as a significant problem within the Groningen Field, located in the northeast corner of the Netherlands. A key concern is the location, source mechanism, and magnitude of the microearthquake events, which are triggered by stress changes induced in the reservoir through pressure depletion. The resulting seismic energy release has been linked to building damage and social unrest in the area. The situation is compounded by the structural complexity of the subsurface with many hundreds of faults that crisscross the area, making earthquake location estimation challenging. An accurate methodology to determine both location and seismic mechanism is required. We present the results of an integrated full-waveform, 3D elastic, event location, and moment-tensor inversion workflow that has been applied to the regional, shallow borehole-monitoring array. This workflow utilizes all event arrivals to improve the accuracy of locating earthquakes and offers a reliable estimate of the earth's sense of motion during an earthquake. It is also demonstrated that the biggest source of error in locating earthquakes at Groningen lies in the vertical depth accuracy, which is correlated to the azimuthal angular coverage, and the fact that small earthquakes are only detected by a relatively small number of stations.
29 May 2023 in Communications Earth & Environment
Moderate to strong earthquakes have been induced worldwide by shale gas development, however, it is still unclear what factors control their behaviors. Here we use local seismic networks to reliably determine the source attributes of dozens of M > 3 earthquakes and obtain a high-resolution shear-wave velocity model using ambient noise tomography. These earthquakes are found to occur close to the target shale formations in depth and along high seismic velocity boundaries. The magnitudes and co-seismic slip distributions of the 2018 Xingwen $${M}_{{{{{{\rm{L}}}}}}}5.7$$ and 2019 Gongxian $${M}_{{{{{{\rm{L}}}}}}}5.3$$ earthquakes are further determined jointly by seismic waveforms and InSAR data, and the co-seismic slips of these two earthquakes correlate with high seismic velocity zones along the fault planes. Thus, the distribution of high velocity zones near the target shale formations, together with the stress state modulated by hydraulic fracturing controls induced earthquake behaviors and is critical for understanding the seismic potentials of hydraulic fracturing.
19 July 2022 in Seismological Research Letters
In recent years, a variety of deep learning (DL) models for seismic phase picking have attracted considerable attention and are widely adopted in many earthquake monitoring projects. However, most current DL models pick P and S arrivals trace by trace without simultaneously considering the spatial coherence of seismic phases among different stations in a seismic array. In this study, we develop a generalized neural network named CubeNet based on 3D U‐Net to properly consider the spatial correlation of individual picks at different stations and thus improve the picking accuracy. To deal with data acquired by irregularly distributed stations, seismic data are first regularized into data cubes, which are then fed into CubeNet to calculate probability distributions of P arrivals, S arrivals, and noise. In addition, a variable trace resampling method for optimizing the differential sampling points between P and S arrivals in a trace for varying array apertures is also proposed to further improve the picking accuracy. CubeNet is trained by 47,000 microseismic data cubes and then tested by three data sets from different arrays with varying apertures and station intervals. It is found that CubeNet is rather resilient to impulsive noise and can avoid misidentifying most of the abnormal picks, which are challenging for the signal‐trace based phase picking methods such as PhaseNet. We believe the newly proposed CubeNet is especially suitable for processing seismic data collected by large‐N arrays.
29 April 2024 in Geophysical Journal International
Investigating subsurface shear wave velocity (vs) structures using surface wave dispersion data involves minimizing a misfit function that is commonly solved through gradient-based optimization. Sensitivity kernels for model updates are commonly estimated using numerical differentiation, variational methods or implicit functions which however, may involve numerical instability and computational challenges when dealing with complex velocity models and large data sets. In this study, we propose a novel surface wave inversion framework in which error-free gradients are calculated by automatic differentiation (AD) and forward modelling is implemented by convenient computational graphs in the state-of-the-art deep learning framework. The AD-based inversion approach is first validated using two synthetic data sets. Then, the subsurface structures at three distinct locations, namely the Great Plains and the Long Beach in the US and Tong Zhou in China, are also derived using this method with seismic ambient noise data, which show nice consistency with those obtained using traditional methods. With the significantly improved computational efficiency, a great number of initial models can be inverted simultaneously to mitigate the impact of local minima and to estimate the uncertainty in the invert models. We have developed a new surface wave inversion package named ADsurf based on automatic differentiation and computational graphs in the deep learning framework, and its computational efficiency is also compared with the traditional finite-difference-based gradient estimation approach. While a great number of intriguing studies on the geophysical inverse problems have been conducted recently using deep learning for end-to-end mapping, the use of AD provided in the in the deep learning frameworks to assist and expedite the gradient computations are still underexploited in geophysics. Thus, it is expected that various geophysical inverse problems in many different areas beyond the surface wave inversion can also be tackled with this new paradigm in the future.
30 March 2022 in Seismological Research Letters
To image high-resolution crustal and upper-mantle structures, ambient noise tomography (ANT) has been widely used on local and regional dense seismic arrays. One of the key steps in ANT is to extract surface-wave group and phase velocity dispersion curves from cross-correlation functions of continuous ambient noise recordings. One traditional way is to manually pick the dispersion curves from dispersion spectrograms in the period-velocity domains, which is very labor intensive and time consuming. Another way is to automatically pick the dispersion curves using some predefined criteria, which are not reliable in many cases especially for phase velocity data. In this study, we propose a novel method named DisperPicker to automatically extract fundamental mode group and phase velocity dispersion curves using a convolutional neural network (CNN). The inputs to CNN include paired group and phase velocity dispersion spectrograms in the period-velocity domains, which are calculated from empirical surface-wave Green’s functions. In this way, group velocity dispersion curves can implicitly guide the extraction of phase velocity dispersion curves, which have large ambiguities to pick on the dispersion spectrograms. The labels or outputs of the network are the probability images converted from dispersion curves. The U-net architecture is adopted because it is powerful for image processing. We have assembled short-period surface-wave data from three different dense seismic arrays to train the network. The trained network is further tested and validated by two datasets close to Chao Lake, China. The tests show that DisperPicker has the generalization ability to efficiently and accurately extract dispersion curves of large datasets without new training.